>>>> Load balancing is at the connection level. So if the users >>>disconnect >>>> and reconnect, they will connect to the correct node. >>> >>>No way to make it spread the load evenly on all nodes? 100% >>>on one node and 0% on the second doesn't strike me as optimal... Um, you don't really want to spread a workload as far an as thin as you can across nodes, do you? Take a workload that could be satisfied by the bandwidth of a single server and throw it out across 2 or 4 and you will quickly learn how much faster local memory is than cache fusion. -- //www.freelists.org/webpage/oracle-l